#AI-RAN #Core Network
Nov 20, 2024
SoftBank Corp.
Blogs
The mobile system has become an indispensable infrastructure in modern life, making its stable operation extremely important. However, upon analyzing the currently mainstream fifth-generation mobile communication system (5G System: 5GS) reveals certain challenges. These include the Core Network (CN), which serves as the control unit, being prone to congestion, and a vulnerability to disruptions in the connection between the base station system (Radio Access Network: RAN) and the CN. While these characteristics may not pose significant issues under normal conditions, there is growing concern that they could lead to severe and prolonged communication disruptions caused by equipment failures, configuration errors, natural disasters, or accidents.
To address these challenges, SoftBank Research Institute of Advanced Technology has proposed the "Autonomous Decentralized Mobile System." In this system, the RAN, integrated with CN functions, is referred to as the Autonomous Decentralized RAN (AD-RAN). The AD-RAN operates autonomously and collaborates in a decentralized manner to maintain the overall functionality of the mobile system. You can read more about the Autonomous Decentralized Mobile System in this article.
Mar 28, 2024
Blogs
Towards More Resilient Mobile Communication
#Core Network
In a mobile system, the CN uses subscriber information (policy data) stored in the subscriber database (Subscriber DB) to authenticate and authorize users, thereby providing communication services. Additionally, the CN performs accounting based on the usage of communication services. Without receiving subscriber information as input, the CN is unable to initiate communication services to the users.
In the Autonomous Decentralized Mobile System, the User Data Repository (UDR), which manages subscriber information, remains centrally located as in traditional CN design. Distributing a database system that handles over 10 million subscriber records in a geographically decentralized manner similar to the RAN is considered unrealistic from both operational convenience and system scalability perspectives. However, relying solely on a centrally located subscriber database, while each AD-RAN is geographically scattered, does not sufficiently account for potential failures in the Transport Network (TN) between AD-RAN and the subscriber database. In such cases, the system cannot be deemed to adequately fulfill the design goal of being resilient to failures. Hence, a highly available subscriber information retrieval system is essential, even while assuming a centrally positioned UDR.
Making changes to the dispersed AD-RANs or the centrally located UDR is not considered advisable. Therefore, SoftBank Research Institute of Advanced Technology considers inserting a layer called the "Replication Layer" between the UDR and AD-RAN to enhance subscriber information availability [2].
Figure 1 illustrates the mechanism of the Replication Layer. All requests sent from AD-RAN to UDR are intercepted by the Replication Layer, which then sends the requests to UDR on behalf of AD-RAN. The Replication Layer stores (replicates) the subscriber information returned from UDR as cached data. Furthermore, the Replication Layer provides the same interface to AD-RAN as UDR. Consequently, when AD-RAN sends a request to UDR, it transparently utilizes the Replication Layer. Since subscriber information once used always passes through the Replication Layer, any subsequent requests for the same subscriber information will hit the cache within the Replication Layer. If a cache hit occurs, the Replication Layer returns the subscriber information to AD-RAN without sending the request to the central UDR.
Figure 1 Cache (replication) of subscriber information by the Replication Layer.
As shown in Figure 2, even if temporary disruptions occur in the communication with UDR or within the UDR itself, accessing cached data within the Replication Layer allows AD-RAN to access subscriber information. Consequently, procedures such as network registration and circuit establishment for User Equipment (UE) can be provided, enabling users to continue accessing communication services.
Figure 2 Behavior of the Replication Layer during disruptions.
Moreover, deploying Replication Layer nodes in dispersed locations allows AD-RAN to utilize the closest nodes. Implementing nodes that contain the Replication Layer closer to AD-RAN can reduce response latency compared to using the centrally located UDR and help distribute the load concentrated on UDR. Therefore, the Replication Layer, operating transparently to both the UDR and AD-RAN, is expected to enhance the availability of subscriber information and contribute to improved fault tolerance across the entire mobile system.
Inserting the Replication Layer does not resolve all issues. Here, we organize emerging discussions due to the Replication Layer handling subscriber information.
When only the centrally located UDR handles subscriber information,the system operates under the assumption of consistent subscriber information. However, when both UDR and the Replication Layer handle subscriber information, the question arises of how much consistency should be maintained. Subscriber information is not only read but also written and updated. Therefore, when subscriber information on the dispersed Replication Layer nodes is updated, it may lead to inconsistencies with the original data stored in the UDR. Since operators rely on the UDR as the primary subscriber database, some degree of discrepancy between the Replication Layer’s cached state and the UDR’s state must be tolerated. Furthermore, the Replication Layer replicates and distributes subscriber information within the network, which traditionally only UDR held, marking a significant deviation from conventional subscriber information operation policies. Hence, enhancing the resilience of mobile systems is a significant theme that impacts the policies of managing subscriber information.
Enhancing the resilience of mobile systems will continue to grow in importance. This time, we designed the Replication Layer to improve the availability of subscriber information by replicating and distributing it within the network.
Future efforts will focus on quantitatively evaluating whether the Replication Layer can handle a scale of 10 million UEs, as envisioned for real-world operations. Additionally, experimental scenarios simulating TN failures will be developed to verify the Replication Layer's response to various failure cases. Going forward, discussions will focus on the availability of not only subscriber information but also data called contexts generated from subscriber information by CN functionalities.
References
[1] SoftBank Research Institute of Advanced Technology, “Towards More Resilient Mobile Communication”,
https://www.softbank.jp/en/corp/technology/research/story-event/046/, Mar. 2024.
[2] Atsuki Takata, Hiroki Watanabe, Keiichi Shima, Horiba Katsuhiro, Satoshi Uda, Yoichi Shinoda, “Distributed AAA in an Autonomous Distributed Mobile System Environment”, IEICE Tech. Rep., vol. 123, no. 397, NS2023-209, pp. 220-223, Feb. 2024.,
https://ken.ieice.org/ken/paper/20240301tcBd/eng/
Writer :Atsuki Takata, Satoshi Uda, Yoichi Shinoda (JAIST：Japan Advanced Institute of Science and Technology), Hiroki Watanabe (SoftBank Corp.)